{"id":"W4393756461","doi":"10.5281/zenodo.6873559","title":"Touché20-Argument-Retrieval-for-Comparative-Questions","year":2020,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Evaluation and Performance Assessment","field":"Decision Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Optech (Canada)","funders":"","keywords":"Argument (complex analysis); Computer science; Information retrieval; Medicine","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002435325,0.004041072,0.001511439,0.003818143,0.001256623,0.002894071,0.004231812,0.004534381,0.06416584],"category_scores_gemma":[0.01293277,0.0008125783,0.002363913,0.003581066,0.0008315353,0.002172596,0.003276142,0.00316442,0.09898188],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00253948,"about_ca_system_score_gemma":0.003641474,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02817369,"about_ca_topic_score_gemma":0.06056552,"domain_scores_codex":[0.9974999,0.000619502,0.0002591985,0.0006372723,0.0007040389,0.0002800087],"domain_scores_gemma":[0.9950228,0.001966412,0.0003452257,0.001210864,0.0009673773,0.0004873533],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0001150089,0.00007635459,0.0006866374,0.0007859374,0.00003375153,0.00003050424,0.00003958263,0.0004658788,0.0002286093,0.0006351282,0.993378,0.003524634],"study_design_scores_gemma":[0.0008283363,0.00007044049,0.005695587,0.0004811531,0.00006695165,0.000126784,0.0002055484,0.002636243,0.00140733,0.003830709,0.9845839,0.00006713727],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.0005182417,0.0001392345,0.0002029655,0.0001480657,0.00004351533,0.00005065362,0.9963744,0.001050937,0.001471981],"genre_scores_gemma":[0.000552417,0.00002740559,0.0006033938,0.00004631607,0.000006218435,0.0001317683,0.9979603,0.00005880799,0.0006132634],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.06416584,"threshold_uncertainty_score":0.2146561,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.27489454462019,"score_gpt":0.4476513850146947,"score_spread":0.1727568403945047,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}